A federated learning data sharing method for Internet of Vehicles

By calculating the similarity and proportion of local and global model parameters and combining Laplace plus noise to optimize the communication process of federated learning in the Internet of Vehicles, the balance between data privacy and efficiency is solved and efficient data sharing is achieved.

CN120546993BActive Publication Date: 2025-10-03QILU INST OF TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511028306.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-07-25
Publication Date
2025-10-03
Estimated Expiration
2045-07-25

AI Technical Summary

Technical Problem

How to improve communication efficiency and reduce computing overhead in federated learning in Internet of Vehicles while ensuring data privacy.

Method used

By calculating the similarity between local model parameters and global model parameters, dynamically updating the similarity of local model participating nodes, and combining the proportion of model parameters in the global model parameters, Laplace plus noise is used to balance data privacy and availability and optimize the communication process.

Benefits of technology

It improves the communication efficiency of Internet of Vehicles data in federated learning, reduces the computational overhead of privacy protection, and provides a balance between data privacy and availability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120546993B_ABST
    Figure CN120546993B_ABST
Patent Text Reader

Abstract

The present application discloses a method for sharing federated learning data for the Internet of Vehicles (IoV), which relates to the field of IoV technology and includes: calculating the parameter similarity between local model parameters and global model parameters; performing local iterative training based on the parameter similarity, and dynamically updating the dynamic similarity of the local model participating nodes; selecting the local model parameters with the best iterative results of the participating nodes to construct a model parameter vector; calculating the weight of each local model parameter in the global model parameter; and performing Laplace noise addition on the global model parameters aggregated by the server based on the model parameter vector and the weight of the local model parameter in the global model parameter. By calculating the dynamic similarity of the local model parameters and their weight in the global model parameters, and by introducing a noise factor to balance the privacy and availability of the data, the communication efficiency of IoV data in federated learning applications is improved, and the computational overhead of privacy protection is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of vehicle networking technology, and specifically to a federated learning data sharing method for vehicle networking. Background Art

[0002] The integration of the Internet of Vehicles (IoV) and information technology has generated massive amounts of data from vehicles, which is crucial for improving traffic efficiency and driving safety. However, the sheer scale and complexity of this data make privacy and security key challenges in IoV data sharing.

[0003] Traditional centralized processing methods pose privacy risks and are unable to meet the data processing needs of large-scale connected vehicle environments. Federated learning, a distributed machine learning framework with privacy-preserving properties, can coordinate thousands of participants in a distributed manner and iteratively train specific machine learning models, providing support and protection for addressing privacy risks in connected vehicles. Participants only need to keep their data locally to collaboratively build a global model. Wang Wenming et al. proposed an authentication and privacy protection method for connected vehicles based on federated learning and drones. The parameters obtained through local training are encrypted using cryptographic techniques such as elliptic curve cryptography and the Chinese remainder theorem. However, given the large scale of data in real life, encrypting all data would incur significant computational overhead. Zhang Ming et al. proposed a secure data sharing and privacy protection method for connected vehicles based on secret sharing. This scheme can apply functions to encrypted data without leaking the data value. However, since the secret share is divided into multiple shares and stored in different locations, this increases overall storage requirements.

[0004] Therefore, how to ensure data privacy while improving communication efficiency in federated learning is an urgent problem that needs to be solved. Summary of the Invention

[0005] In order to solve the above technical problems, this application proposes the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for sharing federated learning data for an Internet of Vehicles (IoV), including:

[0007] Calculate the parameter similarity between the local model parameters and the global model parameters, including:

[0008] Determine the local model parameters respectively and global model parameters ;

[0009] Parameter similarity between local model parameters and global model parameters , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, Indicates the number of parameters;

[0010] Performing local iterative training based on the parameter similarity and dynamically updating the dynamic similarity of the local model participating nodes includes:

[0011]

[0012] Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment.

[0013] Select the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector;

[0014] Calculating the weight of each local model parameter in the global model parameter;

[0015] Laplace noise is performed on the global model parameters obtained by server aggregation based on the model parameter vector and the proportion of the local model parameters in the global model parameters.

[0016] In a possible implementation, selecting the local model parameters with the best iteration results among the participating nodes to construct the model parameter vector includes:

[0017]

[0018] Where: is the model parameter vector, Indicates the local model parameters of the round with the maximum parameter dynamic similarity.

[0019] In a possible implementation, calculating the proportion of each local model parameter in the global model parameter includes:

[0020]

[0021] Where: represents the number of parameters of the local model, represents the importance weight of each local model parameter, represents the number of parameters of the global model, Represents the importance weight of each global model parameter.

[0022] In a possible implementation, performing Laplace noise on the global model parameters aggregated by the server based on the model parameter vector and the proportion of the local model parameters in the global model parameters includes:

[0023]

[0024] Where: is the noise balance factor, , is the global sensitivity, Budget for privacy, Laplacian noise added to the nth vector.

[0025] In a second aspect, an embodiment of the present application provides a federated learning data sharing system for an Internet of Vehicles (IoV), including:

[0026] The similarity calculation module is used to calculate the parameter similarity between the local model parameters and the global model parameters, including:

[0027] Determine the local model parameters respectively and global model parameters ;

[0028] Parameter similarity between local model parameters and global model parameters , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, Indicates the number of parameters;

[0029] A similarity update module is used to perform local iterative training based on the parameter similarity and dynamically update the dynamic similarity of the local model participating nodes, including:

[0030]

[0031] Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment.

[0032] A model parameter vector construction module is used to select the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector;

[0033] A parameter weight calculation module, configured to calculate the weight of each local model parameter in the global model parameter;

[0034] The aggregation module is used to perform Laplace noise on the global model parameters obtained by server aggregation by combining the model parameter vector and the proportion of the local model parameters in the global model parameters.

[0035] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0036] processor;

[0037] Memory;

[0038] And a computer program, wherein the computer program is stored in the memory, and the computer program includes instructions, which, when executed by the processor, enable the electronic device to execute the method described in any possible implementation of the first aspect.

[0039] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the method described in any possible implementation method of the first aspect.

[0040] In an embodiment of the present application, by calculating the dynamic similarity of local model parameters and their proportion in the global model parameters, and by introducing a noise factor to balance the privacy and availability of data, the communication efficiency of Internet of Vehicles data in federated learning applications is improved, and the computational overhead of privacy protection is reduced, providing a new strategy for the application of Internet of Vehicles data in federated learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of a federated learning data sharing method for the Internet of Vehicles provided in an embodiment of the present application;

[0042] Figure 2 A schematic diagram of a federated learning iteration process for the federated learning data sharing method for the Internet of Vehicles provided in an embodiment of the present application;

[0043] Figure 3 A schematic diagram of a federated learning data sharing system for the Internet of Vehicles provided in an embodiment of the present application;

[0044] Figure 4 A schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The present invention will be described below with reference to the accompanying drawings and specific implementation methods.

[0046] See also Figure 1 The federated learning data sharing method for the Internet of Vehicles provided in this embodiment includes:

[0047] S101, calculating the parameter similarity between the local model parameters and the global model parameters.

[0048] Parameter similarity is not affected by the vector modulus, but only focuses on the direction of the vector. It is sensitive to the relative direction and distribution characteristics of the parameters and can well reflect the overall trend and similarity between the parameters. Using the local model parameter vector ( ) and the global model parameter vector ( ) and the size of their respective vectors to calculate the parameter similarity. The calculation formula is as follows:

[0049] , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, represents the local model parameters, represents the global model parameters, Indicates the number of parameters.

[0050] S102: Perform local iterative training based on the parameter similarity to dynamically update the dynamic similarity of the local model participating nodes.

[0051] After every three iterations of local training, the parameter similarity of participating nodes is dynamically updated. The dynamic similarity calculation formula is as follows:

[0052]

[0053] Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment.

[0054] S103, selecting the local model parameters with the best iteration results among the participating nodes to construct a model parameter vector.

[0055] Based on S102, select the model parameters with the best iteration results of the participating nodes, that is, select the model parameters with the largest round of dynamic similarity of parameters, and construct the model parameter vector .

[0056]

[0057] Where: Indicates the model parameters of the round with the maximum parameter dynamic similarity.

[0058] S104: Calculate the weight of each local model parameter in the global model parameter.

[0059] In this embodiment, the proportion of each local model parameter in the global model parameter is calculated. , the specific formula is as follows:

[0060]

[0061] Where: represents the number of parameters of the local model, represents the importance weight of each local model parameter, represents the number of parameters of the global model, Represents the importance weight of each global model parameter.

[0062] S105 , performing Laplace noise addition on the global model parameters obtained by server aggregation based on the model parameter vector and the proportion of the local model parameters in the global model parameters.

[0063] In conjunction with S103, the model parameters of the round with the maximum dynamic parameter similarity are selected and uploaded to the server. The server aggregates the global model to obtain the global model parameters. In conjunction with S104, the global model parameters are perturbed with Laplace differential privacy using the proportion of local parameters in the global model parameters.

[0064]

[0065] Where: is the noise balance factor, , is the global sensitivity, Budget for privacy, The Laplace noise added to the n-th row and n-th column vector parameters. The importance of the model parameters can be obtained by calculating the weight of each local model parameter in the global model parameter formula. , indicating that the parameters are more important for the convergence of the global model, so setting Control the amount of noise added to achieve a balance between data privacy and usability.

[0066] See also Figure 2The federated learning data sharing method for the Internet of Vehicles provided in this embodiment performs a federated learning iterative process. The client performs iterative training locally. The local model parameters obtained through training are screened for parameter similarity. Model parameters with high similarity to global model parameters are uploaded to the server for global model parameter aggregation. After aggregation, they are sent to the local local model to continue iterative training. The above steps are repeated until the model converges or the preset number of iterations is reached.

[0067] Corresponding to the above-mentioned embodiment of a federated learning data sharing method for the Internet of Vehicles, the present application also provides an embodiment of a federated learning data sharing system for the Internet of Vehicles.

[0068] See also Figure 3 The federated learning data sharing system 200 for the Internet of Vehicles in this embodiment includes:

[0069] The similarity calculation module 201 is used to calculate the parameter similarity between the local model parameters and the global model parameters, including:

[0070] Determine the local model parameters respectively and global model parameters ;

[0071] Parameter similarity between local model parameters and global model parameters , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, Indicates the number of parameters.

[0072] The similarity updating module 202 is configured to perform local iterative training based on the parameter similarity and dynamically update the dynamic similarity of the participating nodes of the local model, including:

[0073]

[0074] Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment.

[0075] The model parameter vector construction module 203 is used to select the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector.

[0076] The parameter proportion calculation module 204 is used to calculate the proportion of each local model parameter in the global model parameter.

[0077] The aggregation module 205 is configured to perform Laplace noise addition on the global model parameters obtained by server aggregation by combining the model parameter vector and the proportion of the local model parameters in the global model parameters.

[0078] Corresponding to the above embodiment, an embodiment of the present application further provides an electronic device.

[0079] See also Figure 4 , is a structural diagram of an electronic device provided in an embodiment of the present application. Figure 4 As shown, the electronic device 300 may include: a processor 301, a memory 302, and a communication unit 303. These components communicate via one or more buses. Those skilled in the art will appreciate that the electronic device structure shown in the figure does not limit the embodiments of the present application. It may be a bus structure or a star structure, and may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0080] The communication unit 303 is used to establish a communication channel so that the electronic device can communicate with other devices.

[0081] The processor 301 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. It runs or executes software programs and / or modules stored in the memory 302, and calls data stored in the memory to perform various functions of the electronic device and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 301 can include only a central processing unit (CPU). In the embodiment of the present application, the CPU can be a single computing core or multiple computing cores.

[0082] The memory 302 is used to store execution instructions of the processor 301. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0083] When the execution instructions in the memory 302 are executed by the processor 301 , the electronic device 300 is enabled to execute part or all of the steps in the above method embodiment.

[0084] Corresponding to the above embodiment, embodiments of the present application further provide a computer-readable storage medium, wherein the computer-readable storage medium may store a program. When the program is executed, the program may control the device containing the computer-readable storage medium to execute some or all of the steps of the above method embodiments. In a specific implementation, the computer-readable storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0085] Corresponding to the above embodiment, an embodiment of the present application further provides a computer program product, which includes executable instructions. When the executable instructions are executed on a computer, the computer executes some or all of the steps in the above method embodiment.

[0086] In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0087] The above description is merely a specific embodiment of the present application. Any person skilled in the art may easily conceive of variations or substitutions within the technical scope disclosed in this application, and such variations or substitutions shall be within the scope of protection of this application. The scope of protection of this application shall be subject to the scope of protection of the claims.

Claims

1. A method for sharing federated learning data for Internet of Vehicles, characterized by: include: Calculate the parameter similarity between the local model parameters and the global model parameters, including: Determine the local model parameters respectively and global model parameters ; Parameter similarity between local model parameters and global model parameters , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, Indicates the number of parameters; Performing local iterative training based on the parameter similarity and dynamically updating the dynamic similarity of the local model participating nodes includes: Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment. Select the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector; Calculating the weight of each local model parameter in the global model parameter; Laplace noise is performed on the global model parameters obtained by server aggregation based on the model parameter vector and the proportion of the local model parameters in the global model parameters.

2. The method for sharing federated learning data for Internet of Vehicles according to claim 1, characterized in that: The selecting of the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector includes: Where: is the model parameter vector, Indicates the local model parameters of the round with the maximum parameter dynamic similarity.

3. The method for sharing federated learning data for Internet of Vehicles according to claim 2, characterized in that: The calculating the proportion of each local model parameter in the global model parameter includes: Where: represents the number of parameters of the local model, represents the importance weight of each local model parameter, represents the number of parameters of the global model, Represents the importance weight of each global model parameter.

4. The method for sharing federated learning data for Internet of Vehicles according to claim 3, characterized in that: The performing Laplace noise addition on the global model parameters obtained by server aggregation by combining the model parameter vector and the proportion of the local model parameters in the global model parameters includes: Where: is the noise balance factor, , is the global sensitivity, Budget for privacy, Laplace noise added to the nth row and nth column vector parameter.

5. A federated learning data sharing system for Internet of Vehicles, characterized by: include: The similarity calculation module is used to calculate the parameter similarity between the local model parameters and the global model parameters, including: Determine the local model parameters respectively and global model parameters ; Parameter similarity between local model parameters and global model parameters , where: Indicates parameter similarity, represents the local model parameter vector, represents the global model parameter vector, Indicates the modulus length calculation, Indicates the number of parameters; A similarity update module is used to perform local iterative training based on the parameter similarity and dynamically update the dynamic similarity of the local model participating nodes, including: Where: represents the dynamic similarity, Indicates the time up to the current moment, which is a dynamically changing value. is the time decay function, is the initial value, is the decay rate, The larger it is, the faster it decays. t is the time interval, that is, the time difference from the initial moment to the current moment. A model parameter vector construction module is used to select the local model parameters with the best iteration results of the participating nodes to construct the model parameter vector; A parameter weight calculation module, configured to calculate the weight of each local model parameter in the global model parameter; The aggregation module is used to perform Laplace noise on the global model parameters obtained by server aggregation by combining the model parameter vector and the proportion of the local model parameters in the global model parameters.

6. An electronic device, characterized in that: include: processor; Memory; and a computer program, wherein the computer program is stored in the memory, and the computer program includes instructions, which, when executed by the processor, enable the electronic device to perform the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 4.

Citation Information

Patent Citations

  • Edge-based federated learning model cleaning and equipment clustering method, system and equipment and readable storage medium

    CN112181971A

  • Vehicle selection method and system in Internet of Vehicles federated learning

    CN116546429A